Matthew Di Ferrante

Since I keep having this discussion lately, I wrote an in-depth post about why Quantum Computing will take much longer than many pretend it will, debunking a lot of the current misleading hype/information on the topic: mattdf.xyz/why-quantum-co…

@enggirlfriend 90% of the entire isekai genre is usually about how life in japan is so horrible that it's better to just die and reroll the dice in a new world

i’ve noticed a big difference in japan vs US media culture via anime vs american TV shows: - animes are mostly very pro-japan. japan is always presented in a positive light - animes almost always have societally positive messages like power of friendship, dangerous dystopias, hero’s journey - the story lines are incredibly creative and unique in comparison american tv shows: - overly sexual - shoving DEI down our throats - a lot of “shitting on america” themes - spamming the same template over and over again for views so it’s boring i hope US media can take some learnings and start actually loving ourselves and be creative and positive

@suchenzang why in the world would they choose this as an example

how $125 billion market cap company came to thoroughly enjoy the smell of their own AI garbage must be studied

@_joe_harris_ your dms aren't open, or maybe just closed bc i let my premium lapse

robotics needs better talent, not just ideas or capital get good at any of these and become the person every robotics team is trying to hire: autonomy stack: – state estimation, planning, controls or the in‑house stack nobody else can touch sim & test infrastructure: – lossless logs, reproducible sims, rl loops (nvidia isaac sim, gazebo, mujoco) fleet ops & deployment: – ota updates, connectivity, getting data off robots in the field (greengrass, alloy, formant, or duct tape) data, debugging & replay: – figuring out why the robot did what it did, logs, time‑series, post‑mission analysis (mostly homegrown, rerun/foxglove, alloy) embedded & edge systems: – getting all of this to run on jetson / rb5 / weird industrial pcs safety, compliance & verification: – kill switches, test harnesses, ethics boards, fda submissions, and the standards work nobody wants to do data engine & labelling: – building the labelling, eval, and feedback loops that keep the robot from drifting into chaos go to market & raas: – pricing, contracts, usage‑based billing, customer success for robots‑as‑a‑service if you’re trying to jump into robotics (or want to work with us), my dms are open 🦾

XPeng CEO Questions Claims Around “World Models” in Robotics On May 11, XPeng CEO He Xiaopeng questioned recent claims from some robotics companies about breakthroughs in “physical world models.” He argued that if these systems had truly solved world modeling, they should theoretically be able to generate data autonomously instead of continuously relying on large-scale external data collection. He also suggested that such claims can be tested through practical indicators including data storage scale, R&D cost, and generalization ability without repeated mapping or on-site tuning. Drawing comparisons to autonomous driving, He noted that embodied intelligence generates roughly 10,000x less daily data volume than the smart driving industry, while high-quality autonomous driving datasets have already exceeded 1000PB+ scale. He added that even a relatively small end-to-end autonomous driving model can still require R&D spending of roughly 14 to 70 Million dollars Monthly. According to He, today’s embodied AI sector is beginning to resemble the earlier autonomous driving cycle, where hype driven by concepts and demo videos arrived long before large-scale real-world deployment.

This is like one of those audio journals you find on a corpse in a sci-fi horror game

@johnloeber at least for me memory space is kind of category grouped. autobiographical memory is not overridden by new factual knowledge. but factual confuses factual, e.g., learning a 4th language does make it harder to remember words in the 3rd language i learned.

One thing nobody prepares you for is that professional success means that you dedicate an enormous amount of headspace to things you do not care at all about. It becomes a memory burden. I know so much about taxes. Corporate and personal, US and EU. I wish I didn't. I know so much about corporate law. Endless regulatory nuances in insurance. And at some point you realize that this is actually in zero-sum competition with stuff that you do care about. You can only remember so much. Did I trade off priceless childhood memories for memorizing a bunch of details about Android Development or Biden-Era Crypto Regulation? I sure hope not.

@HyperLogistix @SGhashghaei @HTrobotics i have 2 of these in my lab, they're performing well so far, the software is a bit wonky but that's fixable since it's all open source.

@SGhashghaei @HTrobotics I'm going to wait until I come across someone with one. Lots look good on paper but have fatal flaws.

🚀 Panthera-HT is here — the cost-efficient robotic arm that doesn't compromise performance. Only $1,799.3.5kg payload | 0.8m/s speed | ≤140μs response Fully open-source (MIT) | LeRobot-native | ROS2 ready #Robotics #OpenSource #EmbodiedAI #LeRobot #XLeRobot

@Duduqa1 nope, it's still a theory / computational QC paper, my doubts are in the physical realization. like the impossibility of building real coherent devices anytime soon.

@matthewdif Hi, Matthew! I apologize if I´m wasting your time (I´m a layman in both cryptography and QC), but does this change any of your previous writings? x.com/nic_carter/sta…

i'm gonna laugh if large scale arbitrary quantum computation is actually fundamentally impossible due to entanglement operating in a constrained cohomological physical space, and all the QC hand wringing for the past several decades turns out to have been a giant waste of time

@maximalcoupling it gets so much worse, kms

i should've gone into industry man

wish you could just mute posters from entire cities/regions from appearing in your feed, pretty tired of all the sf griftslop

Has anyone done some proper benchmarking on abliterated models vs their unmolested baseline? In my experience abliteration seems to significantly impact accuracy especially on harder/more fragile tasks, but would be good to see this quantified.

i started learning chinese 1.5 years ago as it became obvious to me then that China was much further ahead in real economic output and tech progression pace than basically everyone else and was only going to keep accelerating one thing i definitely did not predict is that the US' reaction to this would be to immediately light themselves on fire and do everything possible to destroy their unique competitive advantages while also destroying all the "world steward" reputation they had built up over the past century in the process the incompetence of the political class on this side of the world is astounding, you can never underestimate their ability to make the absolute worst possible choices at every turn

@1llegalEngineer a FY6900-100M good enough? if you're close to Switzerland could bring it over to you

Anybody willing to sell a good ish function generator or digital scope? I the emp of the reactor test run fried both.

True Patriots will know Trump is referring to the unresolved contention between magnetic monopoles predicted in high-energy theory and the observed dipoles of the EM field. It's clear he believes in the quantization-algebra view of QFT, in which broken U(1) gauge symmetry in electromagnetism leads to nonlocal single-particle entanglement mediated through Faddeev-Popov ghost dynamics, which implies that the quantization structure itself drives intrinsic field behavior, rather than observable on/off-shell particle interactions.

@rieszspieces this tweet in itself could be a good ragebait if you just quote someone else's math-related tweet with it

you talk about math in a way that makes it clear that you don't know what you're talking about to people that actually know what you're talking about but makes it look like you know what you're talking about to people that don't know what you're talking about

@malikules it seems the issue is his research area is computer vision and has nothing to do with biology lol

it's 2025 and one still has to learn proper form for exercises over months of practice instead of just installing a program into the reflex arc via some neuralink type shit. singularity in 5 years my ass.

An entanglement-enhanced optical clock achieves unprecedented precision of 10⁻¹⁸ with 2dB quantum noise squeezing go.aps.org/4oMKSFD

🚀 Hello, Kimi K2 Thinking! The Open-Source Thinking Agent Model is here. 🔹 SOTA on HLE (44.9%) and BrowseComp (60.2%) 🔹 Executes up to 200 – 300 sequential tool calls without human interference 🔹 Excels in reasoning, agentic search, and coding 🔹 256K context window Built as a thinking agent, K2 Thinking marks our latest efforts in test-time scaling — scaling both thinking tokens and tool-calling turns. K2 Thinking is now live on kimi.com in chat mode, with full agentic mode coming soon. It is also accessible via API. 🔌 API is live: platform.moonshot.ai 🔗 Tech blog: moonshotai.github.io/Kimi-K2/thinki… 🔗 Weights & code: huggingface.co/moonshotai